An experimental and kinetic modeling study of the pyrolysis and oxidation of n-C3C5 aldehydes in shock tubes

An experimental and kinetic modeling study of the pyrolysis

and oxidation of n-C

3

AC

5

aldehydes in shock tubes

Matteo

Pelucchi

a,⇑

,

Kieran P. Somers

b

,

Kenji Yasunaga

c

,

Ultan Burke

b

,

Alessio Frassoldati

a

,

Eliseo Ranzi

a

,

Henry

J. Curran

b

,

Tiziano Faravelli

a

a_{Dipartimento di Chimica, Materiali ed Ingegneria Chimica ‘‘G. Natta’’ Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy} b_{Combustion Chemistry Centre, National University of Ireland, Galway, University Road, Galway, Ireland}

c_{Department of Applied Chemistry, National Defense Academy, Hashirimizu 1-10-20, Yokosuka 239-8686, Japan} Received 21 March 2014

Received in revised form 30 May 2014 Accepted 31 July 2014

Available online 24 August 2014

1. Introduction

The depletion of fossil fuel reserves and the stringent targets for air pollution reduction have largely increased the focus on gaseous, liquid and solid biofuels as a sustainable source of energy for trans-port, domestic and industrial applications. Biomasses can be used to produce either liquid or gaseous biofuels for transportation

purposes (hydrogen, methane, ethanol and long chain alcohol, dimethyl ether, diesel) through different processes such as Bio-mass-To-Liquid (BTL) or as a side product of Gas-To-Liquid (GTL) processes in Fischer–Tropsch synthesis. Similarly, municipal solid waste (MSW), agricultural and forest residues can also be con-verted to feedstock for energy production through gasification or combustion, followed by conventional power generation cycles.

Within this scenario, low and high molecular weight aldehydes are known to be toxic, some of them carcinogenic, and precursors of free radicals leading to the formation of ozone and urban smog ⇑ Corresponding author.

[1]. C4and C5aldehydes belong to the class of non-regulated pollu-tants and they are classified as mobile source air toxic (MSAT) compounds. Recent fundamental studies of biofuel combustion have addressed the strong belief that long-chain alcohols (propa-nol, butanol, pentanol and related isomers) are likely to be used as an alternative to conventional gasoline (n-butanol and iso-pent-anol mainly) either as additives in order to reduce pollution in terms of PAH, particulates and soot formation. Fundamental stud-ies on the oxidation of alcohols identified the presence of alde-hydes as intermediate products derived from radical as well as molecular dehydrogenation reactions [2–15].

Aldehydes are products of partial or incomplete combustion and they are released into the atmosphere from conventional spark ignition (SI) gasoline and compression ignition (CI) diesel engines, and also from biomass gasification or aerobic treatments [16]. Grosjean et al. [17] studied carbonyl emissions from light-duty and heavy-duty vehicles on a motorway tunnel, detecting emis-sions of saturated, unsaturated and aromatic aldehydes. Zervas

[18] analyzed the exhaust gases from a diesel engine recording high emission of carbonyl compounds, particularly when synthetic fuels were tested compared to a commercial fuel. Karavalakis et al.

[19] highlighted how the use of a Euro4 diesel engine increases the emissions (aldehydes and ketones) compared to a Euro3 engine identifying linear aldehydes from formaldehyde up to C5AC6and aromatic aldehydes. De Abrantes and co-workers focused on form-aldehyde and acetform-aldehyde emissions from diesel engines [20], highlighting higher concentrations than those observed in spark ignition engines. Gasoline and diesel engine emissions of alde-hydes were also compared by Roy [21] through high performance liquid chromatography. Significant and comparable amounts of formaldehyde, acetaldehyde and propanal were detected from SI and CI engines investigating different injection technologies. Schauer [22] measured the emissions of aldehydes from a gaso-line-powered vehicle, detecting concentrations in the order of those measured for diesel engines. Other works in recent years focused on the influence of diesel–biodiesel blended fuels on car-bonyl emissions, agreeing that alternative fuel blending increases the release of aldehydes and that the engine technology influences the phenomena as much as the kind of fuel blend itself [23–27].

From a pure chemical kinetics perspective aldehydes are pri-mary stable intermediate products of biofuel oxidation and pyroly-sis, influencing reaction pathways and important combustion properties [28]. Therefore their combustion behavior is non-negli-gible with respect to designing more efficient and environmentally friendly combustion systems. For all of these reasons, the under-standing of combustion kinetics of aldehydes and furthermore a deeper understanding of the reactivity of the carbonyl side of the molecule (RACH@O) under combustion or pyrolytic conditions, play a crucial role in the capability of kinetic mechanisms to better predict pollutant release from both conventional and renewable fuels.

Since the pioneering work of Dean and co-workers [29] and of Hochgreb and Dryer [30], several kinetic studies on the pyrolysis and oxidation of formaldehyde at low- and high-temperature were undertaken [31–38].

Similarly, acetaldehyde oxidation has been described in detail by different authors. Dating back to the early 1970s Halstead et al. [39] discussed cool flames phenomena and low-temperature oxidation. Oscillatory ignitions in the low-temperature regime were studied in the early 1980s by Gray and co-workers [40], and in the 1990s by Cavanagh et al. [41] and by Di Maio et al.[42]. Kaiser et al. proposed a model to describe the negative tem-perature coefficient (NTC) behavior of acetaldehyde [43]. The high-temperature oxidation of acetaldehyde was investigated by Dagaut et al. [44] in a jet-stirred reactor and in a shock tube, while Hidaka and co-workers studied its pyrolysis in a single-pulse shock

tube [45], and Yasunaga et al. [46] suggested a mechanism describ-ing the oxidation and pyrolysis for the high-temperature regime.

Da Silva and Bozzelli [47] calculated the enthalpies of formation of C2to C7n-aldehydes through quantum chemical calculations and determined bond dissociation energies (BDE) for all CAC and CAH bonds in the molecules. Their study revealed that the RACH2 CH@O bond is the weakest bond in all the aldehydes larger than acetaldehyde, due to the formation of the resonantly stabilized vin-oxy radical. Kaiser [48] developed a chemical kinetic model to describe propanal oxidation in the temperature range 400–700 K. An experimental and modeling study at higher temperatures was carried out by Lifshitz and co-workers [49]. They investigated the thermal decomposition of propanal in a single pulse shock-tube under pyrolysis conditions. A sub-mechanism of 52 elementary reaction steps and 22 species was developed, and the kinetic anal-ysis emphasized the importance of unimolecular initiation reac-tions involving a CAC bond breaking in predicting the intermediate and product species. Furthermore, the species profile predictions were found to be sensitive to the ratio between abstraction by ethyl radical on the fuel molecule and ethyl radical decomposition to form ethylene and a hydrogen atom. Kasper et al.[50] studied the combustion chemistry of propanal in a stoichiom-etric flame at low pressures with molecular beam mass spectrom-etry, highlighting, under the investigated conditions, the importance of alkyl radical addition to the fuel molecule and the need of more detailed kinetic studies to assess the relative impor-tance of available reaction pathways. Akih-Kumgeh and Bergthor-son [51] studied the ignition of propanal in a shock tube and developed a mechanisms underlining the importance of the initia-tion reaction forming _C2H5 and H _CO and the H-atom abstraction reactions from the alpha-carbonyl site in predicting ignition delay times. Laminar flame speeds were measured by Veloo et al. [52] together with jet-stirred reactor experiments to quantify reactant, intermediate and product concentrations. A model to describe the low- and high-temperature oxidation of propanal was then devel-oped and validated.

Veloo and co-workers also studied n-butanal and iso-butanal oxidation in laminar flames and in a jet-stirred reactor proposing a comprehensive mechanism for low- and high-temperature com-bustion consisting of 244 species and 1198 elementary reaction steps [53]. The authors highlighted the importance of _C3H5radical distribution and other key reaction subsets of the mechanism. An interesting comparison between n- and iso-butanal and propanal is also shown for flame speed data, with an emphasis on radical species profiles such as H _CO and _OH of high importance in flame propagation phenomena. Davidson and co-workers investigated the ignition behavior of n-butanal [54] as part of an update to the Dooley et al. methyl butanoate mechanism [55]. Zhang mea-sured ignition delay times for n-butanal [56] and i-butanal [57] over a wide range of equivalence ratios, pressures and tempera-tures, and developed kinetic sub-models for both of the C4alde-hydes based on a literature review and validated them against the measured data.

To the best of our knowledge, no experimental measurements currently exist for n-butanal pyrolysis and n-pentanal pyrolysis and oxidation. Furthermore, despite C3AC5 aldehydes being included in many hydrocarbon and bio-fuels kinetic models, no studies have specifically addressed to develop and validate an oxi-dation mechanism for n-pentanal. There have clearly been signifi-cant efforts focused on understanding the combustion behavior of aldehydes. Yet a consistent and detailed summary of their combus-tion behavior is still somewhat lacking in the literature.

Two important goals support and justify the present work. Firstly, it provides new experimental data on the pyrolysis and the auto-ignition behavior of C3AC5 aldehydes, in order to extend the database available for high-temperature combustion

C0AC4 mechanism, which is beyond the aim of this study, and would require more time and efforts. Moreover, the POLIMI kinetic mechanism is systematically reduced by using a lumping approach; this is not considered nor applied in the NUIG mechanism.

Both the mechanisms were previously validated for formalde-hyde and acetaldeformalde-hyde. Figure 1 shows a first comparison with experimental data. Figure 1a shows the oxidation of formaldehyde in an atmospheric pressure flow reactor at 1095 K and

u

= 1.08 [31], while Fig. 1b shows the high-temperature pyrolysis of acetal-dehyde in a shock tube [46]. Figure 2 compares the ignition delay times of formaldehyde [34] and acetaldehyde [46] (1% fuel in oxy-gen-argon) at lean conditions (

u

= 0.5). The higher reactivity of formaldehyde is evident in this figure, mainly at lower tempera-tures. The apparent activation energies are significantly different, in the order of 40 kcal mol1 _{for formaldehyde and close to 60 kcal} mol1 for acetaldehyde. At temperatures higher than 1600 K, both aldehydes have similar reactivity. These comparisons show that both of the models agree with the experimental data, providing a solid basis for the development of kinetic mechanisms for heavier aldehydes.

2. Experimental approach

Two shock tubes were used in this study, the first one at the National Defense Academy in Japan to study the pyrolysis of alde-hydes, the second one, at the National University of Ireland Galway (NUIG), was used to measure ignition delay times of propanal, n-butanal and n-pentanal at different oxidation conditions.

2.1. Pyrolysis

Pyrolysis experiments were performed in a single-pulse magic-hole type shock tube (SPST) [65–68] at 3% fuel in argon (99.9999% pure, Tokyo Koatsu Yamazaki). Samples of propanal and n-pen-tanal 99.0+% and 97.0+% pure respectively, were supplied by Kanto Chemical Co., Inc., n-butanal 98.0+% pure was supplied by Alfa Aesar. The experimental conditions studied are summarized in Table 1.

The shock tube consists of an adjustable driver section 40–140 cm in length, allowing modifications of the heating time, and a test section measuring 3.60 m in length. The two sections are separated by a polyethylene terephthalate (PET) diaphragm which bursts when a needle is dropped. Three pressure transducers are located along the last 30 cm of the test section, shock velocity at the end-wall is determined through linear extrapolation, taking

0,00 0,05 0,10 0,15 0,20 0,25 0,30 O₂ CO2 CO mol [ % ] t [s] CH2O

(a)

(b)

0,000 0,004 0,008 0,012 1000 1100 1200 1300 1400 1500 1600 0,0 0,2 0,4 0,6 0,8 1,0 CO CH₄ CH3CHO C/ C0 [% ] T [K]

Fig. 1. (a) Oxidation of formaldehyde at 1 atm and 1095 K [31]. (b) Pyrolysis of acetaldehyde at 2 atm and 2 ms residence time [46]. Comparison of experimental data (symbols) and predictions of NUIG (dashed lines) and POLIMI (solid lines) kinetic schemes.

6,0 6,5 7,0 7,5 8,0

10 100 1000

CH

₃

CHO

Ignition Delay Time [ms]

10000/T [1/K]

CH

₂

O

1% Fuel-O

₂

/Ar,

φ=0.5,

p=2 atm

Fig. 2. Ignition delay times of formaldehyde [34] and acetaldehyde [46] oxidation. Comparison of experimental data (symbols) and predictions of NUIG (dashed lines) and POLIMI (solid lines) kinetic schemes.

conditions. Secondly, the experimental data, both from the current study and from the literature, have been used to develop and val-idate the sub-mechanism of C3AC5aldehydes, with the objective of better characterizing the role of the acyl group. This validation has been performed by coupling a newly developed aldehyde sub-mechanism with the C0AC4 mechanism of NUIG [58–63] and the one of POLIMI [64].

The meaning of this joint paper between the two research groups is also a first step toward a unified mechanism. The two C0AC4 mechanisms present several differences in the rate con-stants (within their accuracy) and for both of them it is necessary to preserve internal consistency. Internal consistency means that the branching ratio between competitive reactions is more impor-tant than the values of kinetic parameters of single reactions. For this reason, reactions within the C0AC4 mechanism (including C1AC2 aldehydes) have forced some differences between the implementation of the aldehyde sub-mechanism within the two kinetic _{frameworks. As a matter of fact, the concentration of _H, _OH} and other radicals are mainly controlled by two different C0AC4 mechanisms. More than the same kinetic constants (k) it is important to have the same rate parameters (k [R] s-1), that is the reason why there are differences between the two mecha-nisms. The unification process would need to start from the

shock wave attenuation into account. The Gaseq [69] equilibrium program was used to determine the reflected shock parameters from the known initial temperature, mixture pressure and incident shock velocity via the usual one-dimensional equations. The valid-ity of the experimental method used in this study and reiterated in this Section has been also recently proved in the investigation of the pyrolysis of 2,5-dimethylfuran by Somers et al. [70].

The reacted gas mixtures, quenched using the single-pulse method, were extracted into a pre-evacuated vessel (50 cm3₎ through a valve near the end-plate. The reacted gas mixtures were analyzed by three serially connected gas chromatographs equipped with thermal conductivity detectors (TCD) [67,68].

Shimadzu GC-8APT with helium as carrier gas, equipped with different columns, was used for the different analysis in the post-shock mixtures:

– A 2 m column packed with Sebaconitrile and heated to 348 K to detect propanal.

– A 2 m column of Porapak Q coupled with a 2 m column of Unibeads 1S to determine C2 and C3 hydrocarbons.

– A 2 m column of Molecular Sieve 5A at 323 K to determine methane and CO.

According to Hidaka et al. [66,67], an effective reaction time (te) was defined as the time between the heating of the mixture by the reflected shock wave and the time at which the reflected shock pressure had fallen by 20%. Assuming the adiabatic expansion of a non-reactive mixture, the temperature drops by 8.5% from its initial value at te. Given that the single-pulse shock tube has cool-ing rates of 6.6  105 _{K s}1 _{[65], it can be assumed that the reaction} was frozen at te. The validity of the effective heating time and cool-ing rate was previously tested for N2O pyrolysis [65].

The uncertainty in the measured concentration of small hydro-carbons in the post-shock mixtures is less than 2%, except for fuels where the estimate uncertainty is less than 8% for propanal and n-butanal, and less than 28% for n-pentanal. The uncertainty in the reaction time is 5% and in the reflected shock temperature is±1%. Maximum uncertainties in the reflected pressure are±0.5 atm based on the standard deviation (

r

) of our experiments, while those in the residence time are ±0.5 ms.

2.2. Ignition delay times

The shock tube based at NUI Galway, a standard type UV Emis-sion Shock Tube (UEST), was previously described in detail by Smith et al. [71] and will be briefly discussed here. It consists of a test section measuring 6.22 m in length with an internal diame-ter of 10.24 cm and a barrel-shaped driver section measuring 53 cm in length. The two sections are separated by a polycarbonate diaphragm which bursts when forced into contact with a cross-shaped cutter due to the pressure differential. Four pressure trans-ducers located along the last half meter of the test section were used to determine the velocity of the incident shock wave. The velocity at the end-plate was determined via linear extrapolation, taking shock wave attenuation into account. The Gaseq [69] equilibrium program was used to determine the reflected shock

parameters from the known initial temperature, mixture pressure and incident shock velocity via the usual one-dimensional equa-tions. Given the large diameter of the shock tube facility, the boundary layers have limited effect, thus the test conditions are closely predicted by the one-dimensional equations [72].

The pressure at the end-plate was recorded using a pressure transducer (Kistler, model 603B). Light emission at 431 nm from excited CHH

radical was detected through a fused silica window embedded in the end-plate using a photo-detector (Thorlabs Inc. PDA55-EC) and a narrow band-pass filter centered at 430 nm, with a full width half-maximum of 10 nm. The ignition delay time was defined as the time interval between the rise in pressure due to the arrival of the shock wave at the end-plate and the extrapolation of the maximum slope of CHH

emission to the zero signal level as reported in Fig. 3.

Propanal (97.0+% pure) was supplied by Sigma–Aldrich while n-butanal (99.0%) and n-pentanal (97.0+%) were supplied by TCI Eur-ope. Fuels were degassed before being introduced into the mixing tanks by a series of freeze–pump-thaw cycles until no gas was observed to escape from the fuel as the solid thawed. Oxygen (99.5%) and argon (99.998%) were provided by BOC Ltd and were introduced in the mixing tanks through the manifolds from gas cyl-inders. For each aldehyde, 1% fuel mixtures were prepared at three different equivalence ratios (

u

= 0.5, 1.0 and 2.0) using the method of partial pressures and their compositions are provided in Table 2, together with the temperatures and pressures measured for each mixture. We estimate an uncertainty of ±0.10 and ±0.25 atm in the reflected shock pressure, respectively for 1 and 3 atm experiments, based on the standard deviation (

r

) of the data. Uncertain-ties of 1% are present in the reflected shock temperature. An uncertainty of 15% is estimated in the ignition delay time of each experiment due to uncertainties in the condition behind the reflected shock wave. Uncertainties in the mole fractions of reac-tants are minimal (<5%) as high accuracy digital pressure gauges were used in the preparation of the mixture.

3. Kinetic mechanism of aldehydes 3.1. Thermochemistry

The thermodynamic data for the three aldehydes (Fig. 4) and related radicals, were calculated using the THERM program from

Table 1

Summary of experimental work carried out in the SPST at the National Defense Academy, Japan.

Aldehyde Fuel (%) Ar (%) Trange(K) p (atm) te(ms) Propanal 3.0 97.0 1013–1356 1.8 ± 0.3 2.6 ± 0.5 n-Butanal 3.0 97.0 1096–1368 1.9 ± 0.5 1.9 ± 0.5 n-Pentanal 3.0 97.0 972–1372 2.1 ± 0.4 2.3 ± 0.5

Fig. 3. End-wall pressure trace and CHH

chemiluminescence measurements with corresponding ignition delay time, 1% n-butanal, 5.5% O2, 93.5% Ar, p = 1.2 atm, T = 1349 K.

3.2.1. Unimolecular decomposition reactions

A three-frequency version of Quantum-Rice–Ramsperger–Kas-sel theory (QRRK/MSC) [75–77] was used to calculate the temper-ature and pressure dependency of unimolecular decomposition reactions involving the three aldehydes. Collisional stabilization was calculated using a modified strong collision approximation. The high-pressure limiting rate constants were calculated through microscopic reversibility using estimates for radical–radical recombination reactions. Table 4 shows the resulting BDE com-pared to those evaluated by da Silva and Bozzelli [47]. The CbACc

is the weakest bond, as would be expected from the proximity to the electron withdrawing carbonyl group, followed by CaACb. The weakest CAH bond is the CaAH, followed by CbAH. Once again group additivity methods show good agreement with theoretical computations.

Table 5 reports the high-pressure limit rate parameters of initi-ation reactions in the non-Arrhenius form A Tn _{exp [E}

a/RT], where A is the frequency factor, Ea is the activation energy, and R is the gas constant. Units are: cm3_{, mol, s, cal, K. Relative branching ratios} at four different temperatures of 1000, 1500, 2000 and 2500 K are also shown. Whilst for propanal the decomposition is mainly initi-ated by CaACb breaking throughout the temperature range explored, for T < 1500 K, the chain initiation for butanal and

n-pentanal mainly occurs via the breakage of the CbACcbond to form

the _CH2CHO radical. As already mentioned, the CaAH bond

cleav-age contributes to the chain initiation only at high temperatures, and even then, it is still negligible due to its high activation energy. In order to support the QRRK/MSC approach, a direct compari-son between QRRK/MSC and QC/RRKM/ME approaches for prop-anal

dominant unimolecular decomposition channels (CaACband CbACcbreaking) is presented in Fig. 8. A complete description of

the quantum chemical, RRKM/ME and QRRK/MSC methods and results is provided in Supplementary Material for brevity in the main text. Quantum chemical calculations were carried out using the Gaussian 09 application [78], with the B3LYP functional [79,80] and the CBSB7 basis set used to optimize geometries, to determine frequencies (scaled by 0.99), and to carry out relaxed scans of internal rotors for use in a 1-D hindered rotor approxima-tion. The average energy transferred in a deactivating collision was estimated as hDEdi(T) = 200(T/300)0.85cm1 which is in line with recent RRKM/ME calculations carried out on the potential energy surfaces of propoxy [81] and butoxy [82] radicals. All calculations were carried out in an argon bath gas with Lennard-Jones param-eters of

r

= 3.53 Å and

e

/kB= 162 K assumed. Lennard-Jones

Table 2

Summary of experimental work carried out in the low pressure shock tube at the National University of Ireland, Galway.

Fuel u Fuel (%) O2(%) Ar (%) T5(K) p (atm)

Propanal 0.5 1 8 91 1136–1405 1.25 ± 0.08 1170–1336 2.49 ± 0.13 1 1 4 95 1171–1612 1.22 ± 0.12 1201–1479 2.75 ± 0.25 2 1 2 97 1358–1747 1.21 ± 0.09 1302–1598 2.91 ± 0.30 n-Butanal 0.5 1 11 88 1232–1495 1.10 ± 0.06 1224–1424 3.14 ± 0.11 1 1 5.5 93.5 1303–1547 1.20 ± 0.13 1271–1475 3.23 ± 0.16 2 1 2.75 96.25 1376–1740 1.07 ± 0.05 1321–1610 2.84 ± 0.17 n-Pentanal 0.5 1 14 85 1211–1447 0.98 ± 0.05 1167–1338 3.01 ± 0.33 1 1 7 92 1226–1538 1.06 ± 0.05 1181–1481 2.89 ± 0.15 2 1 3.5 95.5 1363–1847 1.08 ± 0.07 1238–1592 3.16 ± 0.11

Fig. 4. Propanal, n-butanal and n-pentanal chemical structure and named carbon sites.

Ritter and Bozzelli [73], based on group additivity methods developed by Benson [75] and further optimized by Burke [61] at NUIG. The computed values of enthalpies, entropies of formation and heat capacities for aldehydes and primary radicals are shown in Table 3, together with enthalpies of formation computed by da Silva and Bozzelli [47], and the nomenclature used in this study. Good agreement between the group additivity rules and the theo-retical computations is shown in Table 3, with maximum devia-tions being within 1.5–2.0 kcal mol1 _{for the enthalpies of} formation. The bond dissociation energies (BDE) for the three alde-hydes will be analyzed in the next paragraph when discussing chain initiation reactions.

3.2. Primary reactions of aldehydes

Figs. 5–7 show simplified primary chain initiation and propaga-tion reacpropaga-tions of the three aldehydes, in terms of initiapropaga-tion, H-atom abstraction, and radical decomposition reactions. The chain initia-tion reacinitia-tions occur via unimolecular decomposiinitia-tions with a CAC bond cleavage, forming an alkyl radical and an oxygenated radical (H _CO, _CH2CHO, _C2H4CHO-2, _C3H6CHO-3). In flame conditions, the chain initiation and the reverse recombination reactions involving the CAH bonds can become significant. H-atom abstraction reac-tions are reported in their general form with R_ being the generic H-atom abstracting radical. These reactions lead to three, four and five primary fuel radicals, respectively, for propanal, butanal and n-pentanal. Radical decomposition reactions proceed to form either an alkene and an oxygenated radical or an unsaturated oxy-genated species (CO, ketene, acrolein, methylketene, or 1-butenal) and a small alkyl radical. For instance, the

a

-radical derived from the Cn aldehyde could, in principle, decompose via b-scission to form either ketene and a Cn2 alkyl radical or, more likely, CO and a Cn1 alkyl radical. Radicals can also isomerize mainly to form the thermodynamically favored

a

-radical. For the sake of clarity, isomerization pathways are not reported in Figs. 5–7.

parameters for all three aldehydes were estimated from their crit-ical constants [83] and the empirical correlations recommended by Kee and co-workers [84]. RRKM/ME computations for the two dominant unimolecular fission reactions were carried out using the MultiWell code [85,86] using the ILT method to computed k(E). Identical high-pressure limiting rate constants and energy transfer parameters were assumed in QRRK/MSC and RRKM/ME computations. QRRK/MSC results are shown to agree with the more rigorous RRKM/ME computations to within 80% under the conditions tested. Both approaches highlight that above 1 atm, fall-off in the primary unimolecular decomposition pathways of

propanal is limited below 1500 K. Above this temperature, the kinetic model showed little sensitivity to the inclusion of fall-off. As part of this work, the QRRK/MSC approach was also validated against RRKM/ME and experimental measurements of analogous alkane decompositions from Oehlschlaeger et al. [87,88] where a factor of 2–3 agreement in k(T,p) was observed when the QRRK/ MSC and RRKM/ME/experimental recommendations were com-pared. Detailed results together with a detailed quantification of the fall-off behavior of aldehydes are reported in the Supplemen-tary Material attached to this study. The above results re-enforce the applicability of the QRRK/MSC method as a means to include

Table 3

Thermochemical data and nomenclature of aldehydes and related primary radicals (*

group additivity). Aldehyde and radical site Name DH°f(298.15 K) [47](kcal mol1 ) DH°f(298.15 K)* (kcal mol1 ) S°(298.15 K)* (kcal mol1 ) Cp(cal mol1K1)* 300 K 500 K 800 K 1000 K 1500 K Propanal C2H5CHO 45.18 45.35 72.86 19.38 26.94 36.90 42.12 49.83 a C2H5CO_ 8.00 8.45 73.98 18.55 24.98 33.74 38.39 45.19 b _C2H4CHO-1 7.10 5.55 70.49 18.02 25.21 34.24 39.01 46.33 c _C2H4CHO-2 5.10 3.73 76.20 18.85 25.19 33.53 38.03 44.79 n-Butanal n-C3H7CHO 50.00 50.31 82.28 24.82 35.26 48.21 54.29 64.57 a n-C3H7CO_ 13.00 13.41 83.40 23.99 33.30 45.05 50.56 59.93 b _C3H6CHO-1 11.50 10.51 79.91 23.46 33.53 45.55 51.18 61.07 c _{_C3H6CHO-2} 2.50 4.48 86.29 23.29 32.26 44.04 49.65 59.28 d _C3H6CHO-3 0.10 1.23 85.62 24.29 33.51 44.84 50.20 59.53 n-Pentanal n-C4H9CHO 54.61 55.27 91.70 30.26 43.58 59.52 66.46 79.31 a _n-C4H9CO_ 18.30 18.37 92.82 29.43 41.62 56.36 62.73 74.67 b _C4H8CHO-1 17.20 15.47 89.33 28.90 41.85 56.86 63.35 75.81 c _{_C4H8CHO-2} 8.10 9.44 95.71 28.73 40.58 55.35 61.82 74.02 d _C4H8CHO-3 7.60 9.44 95.71 28.73 40.58 55.35 61.82 74.02 e _{_C4H8CHO-4} 5.20 6.19 95.04 29.73 41.83 56.15 62.37 74.27

Fig. 5. Primary decomposition reactions of propanal.

a cost-effective assessment of the influence of fall-off in the kinetic modeling of our high-temperature experiments.

3.2.2. H-abstraction reactions

Rate constant for abstractions of the acyl H-atom at the

a

posi-tion were estimated by analogy with the same site in formalde-hyde and acetaldeformalde-hyde. The presence of the HAC@O groups leads to the formation of resonantly stabilized radicals and the reactivity

of the H-atoms in the b site is slightly enhanced with respect to secondary H-atoms in alkanes. This fact is also evident observing the enthalpy of formation in Table 3 and it is further supported by the kinetic study of 3-pentanone oxidation by Serinyel et al. [89]. Rate constants for abstractions from the remaining secondary and primary H-atoms were adopted according to their values used for n-alkanes [90,91].

Rate constants for this class of reactions need to be defined for all the H-atom abstracting radicals. To maintain an internal consis-tency inside NUIG and POLIMI mechanisms, rate parameters for H-atom abstraction reactions are defined in two different ways, but still within their kinetic uncertainty, as will be discussed in Sec-tions 4.1 and 4.2.

3.2.3. Radical decomposition reactions

Arrhenius parameters for the decomposition of

a

-radicals to form CO were taken from the evaluation by Simmie [92] for 1-oxo-butyl radical (n-C3H7_CO) decomposition to n-propyl radical and CO. b-scission to form ketene and Cn2alkyl radical has also been included in the kinetic scheme, based on the following

Fig. 7. Primary decomposition reactions of n-pentanal.

Table 4

Calculated bond dissociation energies (kcal mol1_{) of CAC and CAH bonds and} comparison to ab initio computed values by da Silva and Bozzelli [47].

Bond Propanal n-Butanal n-Pentanal

This work [47] This work [47] This work [47]

CaACb 84.4 83.8 84.6 84.1 84.5 83.4 CbACc 83.5 83.7 82.3 82.5 82.5 82.3 CcACd 89.1 90.1 87.9 88.5 CdACe 89.1 89.4 CaAH 89.0 89.3 89.0 89.1 89.0 88.8 CbAH 91.9 90.2 91.9 90.6 91.9 89.5 Table 5

High-pressure limit rate parameters of initiation reactions and relative branching ratios at different temperatures [units are: cm3_{, mol, s, cal].}

Reactions ki(s1) Branching ratios (%)

Ai ni Eai 1000 K 1500 K 2000 K 2500 K

Propanal

C2H5CHO M _CH3+ _CH2CHO 1.16E+25 2.80 85718 32.9 31.4 29.7 27.1

C2H5CHO M _C2H5+ H _CO 1.34E+26 3.00 86406 66.5 65.6 62.1 56.0

C2H5CHO M C2H5_CO+ _H 9.42E+16 0.43 89167 0.6 2.7 6.6 12.2

C2H5CHO M _C2H4CHO-1+ _H 1.22E+15 0.08 91694 0.0 0.2 0.7 1.5

C2H5CHO M _C2H4CHO-2+ _H 5.80E+17 0.52 101476 0.0 0.1 0.9 3.2

n-Butanal

n-C3H7CHO M _CH3+ _C2H4CHO-2 1.09E+24 2.25 90369 4.1 14.4 26.2 36.2

n-C3H7CHO M _C2H5+ _CH2CHO 5.04E+27 3.50 84479 67.3 52.3 40.5 31.5

n-C3H7CHO M n _C3H7+ H _CO 7.49E+27 3.51 86758 28.4 32.1 30.0 26.1

n-C3H7CHO M n-C3H7_CO+ _H 2.72E+17 0.58 88995 0.2 1.1 3.0 5.5

n-C3H7CHO M _C3H6CHO-1+ _H 3.62E+15 0.23 91529 0.0 0.1 0.3 0.7

n-Pentanal

n-C4H9CHO M _CH3+ _C3H6CHO-3 1.18E+22 1.61 90120 3.3 9.9 16.5 22.0

n-C4H9CHO M _C2H5+ _C2H4CHO-2 2.84E+24 2.24 89004 17.1 33.4 42.1 45.8

n-C4H9CHO M n _C3H7+ _CH2CHO 1.70E+27 3.31 84704 57.9 35.6 23.1 16.0

n-C4H9CHO M n _C4H9+ H _CO 1.20E+26 2.94 86380 21.6 20.4 16.8 13.7

n-C4H9CHO M n-C4H9_CO+ _H 1.64E+17 0.50 89262 0.1 0.6 1.3 2.1

[93] for the reverse addition reaction are used. Rate parameters for the dehydrogenation reactions of _C2H4CHO-1 radical to form acrolein or methylketene are derived from the kinetic values of the reverse _H-atom addition to propylene to give iso-propyl radi-cal. Methyl and ethyl radical addition to propylene to form the

sec-ondary radicals were used to derive the corresponding

decomposition of b-radicals of n-butanal and n-pentanal, forming acrolein and a methyl or an ethyl radical, respectively.

b-Scission of

c

-radicals to produce formyl radicals and alkenes, were based on computations of n-butanal radical decomposition by Huynh and Violi [94]. Rate constant for the decomposition of propanal

c

-radical to form H _CO and ethylene, was taken from H _CO addition to propylene [94] accounting for the symmetry effect. The

c

-radical of pentanal ( _C4H8CHO-2) can also form a methyl radical and 1-butenal. Similarly to the estimate for methyl radical addition to acrolein in butanal, methyl radical addition to propylene was used as a reference reaction [93].

Again, the n-butanal calculations by Simmie [92] were adopted for the decomposition of d-radicals to form ethylene (or propylene) and _CH2CHO radical.

Finally, the decomposition of the

e

-radical ( _C4H8CHO-4) of n-pentanal to produce ethylene and the

c

-radical of propanal were based on analogy with the addition of alkyl radicals to ethylene, as suggested by Orme et al. [90].

Table 6 summarizes the decomposition reactions and relating kinetic parameters, together with literature references.

3.2.4. Radical isomerization reactions

Radical isomerization reactions, i.e. internal H-atom shift through 3-, 4-, 5- and 6-membered ring intermediates, were also considered. Reaction rate constants were estimated according gen-eral rules [95–97]. The activation energy was estimated through Evans–Polanyi correlations accounting for reaction enthalpy of the H-atom shift reaction and the ring strain energy associated with ring formation in the transition state. Frequency factors were

estimated on the basis of hindered rotor effects. As the ring inter-mediate gets bigger, the reactions become energetically more favored due to lower ring strain energies, but conversely are entro-pically inhibited due to the loss of internal rotors. Isomerization reactions of the different radicals and kinetic parameters are reported in Table 6.

Due to their possible competing effect, a comparison of decom-position and isomerization rate constants of each radical is rele-vant. Figure 9 shows this comparison for n-pentanal radicals in the temperature range 1000–2500 K. Figure 9d shows the 1–5 isomerization reaction of the

e

-radical, occurring through the ener-getically favored six membered ring intermediate, which prevails over the decomposition channel at temperatures lower than 1430 K. The remaining figures show that generally, the decompo-sition paths dominate. Moreover, Fig. 9b shows that for n-pentanal

c

-radical ( _C4H8CHO-2) the decomposition reaction to form 3-bute-nal and a methyl radical prevails over the alternate channel gener-ating 1-butene and a formyl radical. Decomposition of the

a

-radical (1-oxo-pentyl radical, n-C4H9_CO) to form CO and an n-butyl radical largely prevails over the 1–5 isomerization reaction in the complete range of temperature. For this reason reaction rate con-stants for the

a

-radical are not reported in Fig. 9. A similar analysis for propanal and n-butanal radicals is reported in the Supplemen-tary Material.

4. Numerical methods and overall kinetic mechanisms All numerical simulations were performed using the Open-SMOKE code, which is an upgraded and extended version of the well-tested DSMOKE code. A shock tube is considered a constant

1000 1250 1500 1750 2000 10-3 10-2 10-1 100 101 102 103 104 105 106 107 108 1000 1250 1500 1750 2000 k / s -1 T / K k8 10 atm 1 atm (b) T / K (a)

Fig. 8. Comparison of RRKM/ME (symbols) and QRRK/MSC (dashed lines) derived k(T,p) for propanal thermal decomposition to (a) CH3 and CH2CHO and (b) HCO + C2H5. 80% uncertainty bars illustrate the agreement between the two methods.

rationale. Methyl radical addition to ketene to form 1-oxo-propyl radical (C2H5_CO) was taken as a reference and the kinetic parame-ters were estimated by analogy with methyl radical addition to propylene [93]. Rate constants for the analogous b-scission of 1-oxo-butyl and 1-oxo-pentyl radical (n-C4H9 _CO) were based on the ab initio calculations of Huynh and Violi [94].

For the b-radical decomposition to form acrolein (C2H3CHO, 1-propenal) and an alkyl radical the recommended values of Curran

Table 6

Rate constant parameters of decomposition and isomerization reactions of aldehydes radicals [units are: cm3

, mol, s, cal] and references. Reaction

Propanal radicals A n Ea [Ref.]

_C2H4CHO-2 , C2H5CO_ 3.80E+10 0.67 30200.0 [95] _C2H4CHO-1 , C2H5CO_ 3.56E+10 0.88 37300.0 [95] C2H5CO , CO + _C_ 2H5 5.78E+14 0.00 16843.5 [92] a _CH3+ CH2CO , C2H5_CO 1.76E+04 2.48 6130.0 [93] a _

H + C2H3CHO , _C2H4CHO-1 4.24E+11 0.51 1230.0 [93] a

_

H + CH3CHCO , _C2H4CHO-1 4.24E+11 0.51 1230.0 [93] a

H _CO + C2H4, _C2H4CHO-2 2.56E+02 2.89 6728.4 [94] a

n-Butanal radicals

_C3H6CHO-3 , nC3H7_CO 7.73E+11 0.00 15464.4 [95] _C3H6CHO-2 , nC3H7_CO 3.80E+10 0.67 32100.0 [95] _C3H6CHO-1 , nC3H7_CO 3.56E+10 0.88 37300.0 [95] nC3H7_CO , n _C3H7+ CO 5.78E+14 0.00 16843.5 [92] _C2H5+ CH2CO , nC3H7_CO 3.33E+02 2.73 13953.9 [94] _CH3+ C2H3CHO , _C3H6CHO-1 1.76E+04 2.48 6130.0 [93] a

H _CO + C3H6, _C3H6CHO-2 1.28E+02 2.89 6728.4 [94] _C3H6CHO-3 , _CH2CHO + C2H4 3.95E+13 0.00 22316.3 [92] n-Pentanal radicals

_C4H8CHO-4 , nC4H9_CO 3.67E+12 0.60 7090.0 [95] _C4H8CHO-3 , nC4H9_CO 7.85E+11 0.12 16800.0 [95] _C4H8CHO-2 , nC4H9_CO 3.80E+10 0.67 32100.0 [95] _C4H8CHO-1 , nC4H9_CO 3.56E+10 0.88 37300.0 [95] nC4H9_CO , p _C4H9+ CO 5.78E+14 0.00 16843.5 [92] a

n _C3H7+ CH2CO , nC4H9_CO 3.33E+02 2.73 13953.9 [94] a

_C2H5+ C2H3CHO , _C4H8CHO-1 8.80E+03 2.48 6130.0 [93] a

H _CO + C4H8-1 , _C4H8CHO-2 1.28E+02 2.89 6728.4 [94] _CH3+ C3H5CHO , _C4H8CHO-2 1.76E+04 2.48 6130.0 [93] a

_CH2CHO + C3H6, _C4H8CHO-3 1.88E+02 3.11 3660.0 [92] a

_C2H4CHO-2 + C2H4, _C4H8CHO-4 1.32E+04 2.48 6130.0 [90] a

a

volume batch system where the energy equation is solved, thus accounting for both endothermic and/or exothermic gradients. The conservation equations with proper boundary conditions were discretized by means of conventional finite differencing techniques with non-uniform mesh spacing for the simulation of premixed laminar flames. A mixture-averaged formula was also used to com-pute multicomponent diffusion coefficients. Further details of the numerical methods are reported elsewhere [98–100].

The computed sensitivity coefficient, Sy, is normalized (sy) as follows: sy¼dln y dln A¼ A y dy dA¼ A ySy

where y is the species concentration or the ignition delay time and A the generic frequency factor.

The Transport Data Estimator package of the Reaction Mecha-nism Generator software of Green and co-workers has been used to provide relevant transport properties [70].

4.1. NUIG kinetic mechanism

The aldehyde mechanism was coupled with the NUIG C0AC4 sub-mechanism, recently revised and validated as reported in sev-eral recent studies [58–63].

H-atom abstraction rates at the b secondary site forming _CnH2nCHO  1 ðn ¼ 2; 3; 4Þ radicals were based on those for sec-ondary atom abstractions in ketones. Rate constants for H-abstractions by _OH and HO_ 2radicals were taken from the ab initio

calculations for ethylmethyl ketone by Zhou et al. [101] and by Mendes et al. [102] respectively.

Before coupling the aldehyde sub-mechanism with the NUIG kinetic scheme, some modifications were made.

The total initiation rate constant for n-butanal has been increased of a factor of 1.5 with respect to the calculated reaction rates previously discussed in Section 3.2.1. This correction factor, well within the kinetic uncertainty, has been also applied to POLIMI mechanism. H-atom abstraction from the

a

carbonyl site has been decreased by about 20% with respect to the reference abstraction from acetaldehyde. Rate constants for abstraction reac-tions by _H, _OH, _CH3 and HO_ 2radicals are listed in the Supplemen-tary Material. Moreover, overall rate constants of H-abstraction by _

OH radical for C1AC4 aldehydes were studied in shock tubes very recently by Wang et al. [103], further supporting the kinetic parameters here proposed.

The overall kinetic model, with thermodynamic and transport properties, consisting of 2,011 reactions and 329 species is avail-able as part of the Supplementary Material.

4.2. POLIMI kinetic mechanism

Metathesis reactions are treated according to the systematic approach described by Ranzi et al. [104]. The reactivity of primary and secondary H-atoms are considered to be the same as those for alkanes. Similar to the acyl H-atoms of formaldehyde and acetalde-hyde, H-atom abstraction on the

a

site requires a correction of – 4500 cal mol1 _{to be applied to the activation energy for H-atom} abstraction of a primary H-atom from a methyl group. Moreover,

0,0004 0,0006 0,0008 0,0010 105 106 107 108 109 1010 1011 0,0004 0,0006 0,0008 0,0010 105 106 107 108 109 1010 1011 0,0004 0,0006 0,0008 0,0010 107 108 109 1010 1011 1012 0,0004 0,0006 0,0008 0,0010 107 108 109 1010 1011 kde c/ is o m (s -1 ) C₄H₈CHO-1 <=> n-C₄H₉CO C₄H₈CHO-1<=> C₂H₃CHO +C₂H₅

(a)

3 membered ring 4 membered ring

(b)

C₄H₈CHO-2 <=> n-C₄H₉CO C₄H₈CHO-2 <=> 1-C₄H₈+HCO C 4H8CHO-2 <=> C3H5CHO+CH3 5 membered ring

(c)

kde c/ is o m (s -1 ) 1/T (1/K) C₄H₈CHO-3 <=> n-C₄H₉CO C 4H8CHO-3 <=> C3H6+CH2CHO 6 membered ring

(d)

1/T (1/K) C₄H₈CHO-4 <=> n-C₄H₉CO C 4H8CHO-4 <=> C2H4+C2H4CHO-2

Fig. 9. Isomerization (solid lines) and decomposition (dashed lines) reactions of n-pentanal radicals. (a) b-Radical (C4H8CHO-1), (b)c-radical (C4H8CHO-2), (c) d-radical (C4H8CHO-3), (d)e-radical (C4H8CHO-4).

a greater selectivity of H_ and HO_ 2 radicals with respect to the cor-responding one of _OH and _CH3 radicals is also accounted. With regards to abstraction from the b-site, leading to resonantly stabi-lized radicals, a correction factor of about 1.25 has been applied to increase the frequency factor of the secondary H-atom abstraction in alkanes. Rate constants for abstraction reactions by _H, _OH, _CH3 and H _O2radicals are listed in the Supplementary Material. Again, the kinetic parameters here proposed for H-abstractions by _OH radical are further supported by the recent study by Wang et al. [103].

Figure 9 shows that the lifetime of large radicals is so short at high temperatures that they decompose and isomerize without significant interactions with the remaining mixture. When parallel competing reactions are absent or not strongly dependent on tem-perature, larger radicals can be conveniently and directly substi-tuted by their reaction products. Therefore, it is possible to assume them as being directly transformed into their products which are already part to the C0AC4 mechanism [105]. This is the advantage of a lumped approach: it leads to a reduction of the total number of species needed to describe the overall oxidation pro-cess. The analysis of Fig. 9 highlights that only the

e

-radical ( _C4H8-CHO-4) of pentanal shows a temperature dependent competition between isomerization and decomposition rates at high-tempera-tures. Therefore, it is convenient to keep the

e

-radical inside the scheme. The four remaining radicals have been assumed to directly transform into their final products. The same assumption has been also applied to the intermediate radicals of n-butanal and propanal. The effect of these simplifications has proven to be of very limited importance in the high-temperature range of the analyzed condi-tions. At low-temperatures, interactions with oxygen forming per-oxyl radicals will precede the decomposition of primary radicals; therefore, it would be necessary to increase the detail of the involved species.

The oxidation mechanism adopted here [64] consists of over 10,000 reactions and more than 350 species and was developed based on hierarchical modularity. It covers from hydrogen and oxygenated species, up to diesel and biodiesel fuels. The thermo-chemical data for most species in the global mechanism were obtained from the CHEMKIN thermodynamic database [106,107].

For those species whose thermodynamic data are not available in the literature, the group additivity method was used to estimate these properties [75].

The overall kinetic model, with thermo and transport proper-ties, is available in CHEMKIN format from: http://creckmodel-ing.chem.polimi.it and in the Supplementary Material.

5. Model predictions and comparison with experimental data The new experimental data on pyrolysis speciation and ignition delay times for the three aldehydes were used to validate the alde-hyde sub-mechanism in both the kinetic models. Furthermore, propanal pyrolysis data by Lifshitz et al. [49] and ignition delay times by Akih-Kumgeh and Bergthorson [51] were also used. Sim-ilarly, n-butanal ignition delay times measured by Davidson et al. [54] and by Zhang et al. [56] were also compared with model pre-dictions. Finally, comparisons with premixed laminar flame speeds of propanal and n-butanal by Veloo et al. [52,53] complete this kinetic study. The current study, therefore considers all relevant experimental data which exist at present for long chain aldehydes, thus providing a comprehensive evaluation of mechanism perfor-mance. A summary of the experimental data at which the alde-hydes mechanisms were validated in this study is reported in Table 7.

5.1. Pyrolysis in shock tubes

Propanal pyrolysis of a 1% fuel mixture in argon was studied in a single-pulse shock tube by Lifshitz et al. [49] over the temperature range 1000–1300 K. Experimental data and modeling predictions are shown in Fig. 10 for the NUIG and POLIMI mechanisms. Fur-thermore, Fig. 11 compares experimental data from this work and predicted concentration profiles for propanal (3%) decomposi-tion in argon. Propanal undergoes decomposition at 1050 K according to both the experimental data sets and the mechanisms slightly under-predict fuel conversion: an overall good agreement is observed for the main products (CO and C2H4), as well as for the minor hydrocarbon species, particularly in terms of their

Table 7

Experimental data used for the validation of the aldehydes mechanisms.

Aldehyde Reactor/facility T (K) p (atm) U Ref.

Formaldehyde Shock tube Pyrolysis 1160–1890 1.4–2.5 / Hidaka et al.[34]

Shock tube Pyrolysis 1200–2000 1.3–3.0 / Hidaka et al.[36]

Shock tube Pyrolysis 1560–2276 0.9–2.5 / Eiteneer et al.[33]

Shock tube Oxidation 1600–3000 1.0–2.0 0.54, 0.63 Dean et al.[37]

Shock tube Oxidation 1160–1618 1.4–2.5 0.25, 0.5, 1.0, 2.0, 4.0 Hidaka et al.[34] Shock tube Oxidation 1334–1974 0.8–2.3 0.16, 0.25, 1.0, 1.7, 5.9 Eiteneer et al.[33]

Plug flow reactor Oxidation 945 1.0 1.56 Li et al.[38]

Plug flow reactor Oxidation 1095 1.0 0.93 Li et al.[38]

Acetaldehyde Shock tube Pyrolysis 1013–1577 1.2–2.8 / Yasunaga et al.[46]

Jet stirred reactor Oxidation 900–1300 1,10 0.09, 0.43, 0.82, 1.0, 1.61 Dagaut et al.[44]

Shock tube Oxidation 1276–1703 1.7–2.6 0.2, 0.4, 1 Yasunaga et al.[46]

Propanal Shock tube Pyrolysis 970–1300 2.0–2.7 / Lifshitz et al.[49]

Shock tube Pyrolysis 972–1372 1.4–2.8 / This work

Shock tube Oxidation 1170–1750 1.0, 3.0 0.5, 1.0, 2.0 This work

Shock tube Oxidation 1150–1560 1.0, 12.0 0.5, 1.0 Akih–Kumgeh[51]

Premixed flat flame Oxidation 314–2000 0.05 1.0 Kasper et al.[50]

Laminar premixed flames Oxidation 343–2320 1.0 0.75–1.6 Veloo et al.[52]

n-Butanal Shock tube Pyrolysis 1096–1368 1.1–2.8 / This work

Shock tube Oxidation 1190–1550 1.7 1.0, 2.0 Davidson et al.[54]

Shock tube Oxidation 1180–1580 1.3, 5, 10 0.5, 1.0, 2.0 Zhang et al.[56]

Shock tube Oxidation 1224–1634 1.0, 3.0 0.5, 1.0, 2.0 This work

Laminar premixed flames Oxidation 343–2320 1.0 0.75–1.6 Veloo et al.[53]

n-Pentanal Shock tube Pyrolysis 970–1370 1.4–2.8 / This work

0 20 40 60 80 100 0 10 20 30 40 50 1000 1100 1200 1300 0 1 2 3 4 5 1000 1100 1200 1300 0,0 0,1 0,2 0,3 0,4 0,5 C₂H₅CHO CO

C/

C

0

[%

]

C₂H₄ CH₄ C₂H₆ C₂H₂

C/

C

0

[%

]

T (K) C₃H₆ C₃H₈ T (K)

Fig. 10. Predicted and experimental concentration profiles from shock tube pyrolysis of 1% propanal in argon [49] (s = 2.5 ms). Experimental (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

1000 1100 1200 1300 1400 0 20 40 60 80 100 1000 1100 1200 1300 1400 0 1 2 3 4 5 0,0 0,5 1,0 1,5 1000 1100 1200 1300 1400 1000 1100 1200 1300 1400 0 10 20 30 40 50 60 70 C₂H₅CHO CO

C/C

0

[%

]

C₂H₂ C₂H₆ C₃H₈ C₃H₆

C/C

0

[%

]

T (K) C₂H₄ CH₄ T (K)

Fig. 11. Predicted and experimental concentration profiles from shock tube pyrolysis of 3% propanal in argon (s= 2.5 ms). Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

relative concentrations. Analysis of the mechanisms shows that chain radical initiation occurs via unimolecular decomposition reactions involving a CAC bond cleavage, and propanal decomposi-tion mainly occurs via H-atom abstracdecomposi-tion reacdecomposi-tions by H_ atoms and _CH3 radicals.

The POLIMI mechanism under-predicts C2H2 formation reported in Fig. 10 by 40% at temperatures higher than 1200 K and even larger deviations are observed in Fig. 11. The chemistry of the vinyl radical ( _C2H3) is responsible for this under-prediction. H-atom abstractions from acrolein (C2H3CHO) are the main path-ways generating acetylene, either through the vinyl radical formed by the CO elimination of the

a

radical, or via the b-scission of the

c

radical to give formyl radical. The same chemistry is responsible for the over-prediction of acetylene observed for the NUIG mecha-nism in Fig. 10, while good agreement is observed in Fig. 11. The under-predictions of propane and propylene can be attributed to a low methyl radical concentration, i.e. a relatively low importance of the chain initiation reactions. The results shown in Figs. 10 and 11 are considered to be well within the expected experimental uncertainties.

Figure 12 compares experimental and predicted concentration profiles for n-butanal (3%) decomposition in argon. Predictions of fuel conversion as well as the major products, such as CO and eth-ylene, are in very good agreement with the experimental observa-tions. Smaller hydrocarbons detected in the measurements (CH4, C2H2, C2H6, C3H6 and C3H8), are also very well captured by both models. The main relative deviations are observed in the trace amount of propane; this under-prediction is due to methyl and ethyl radical recombination. In analogy with the direct formation of methane and CO from acetaldehyde [46], molecular reactions to form CO or ketene and the corresponding Cn1 or Cn2 alkane are considered, and they play only a limited role under the inves-tigated conditions.

Figure 13 shows a global rate of production analysis performed at 1200 K, for both the NUIG and POLIMI mechanisms. The chain initiation occurs via unimolecular decomposition reactions, the successive decomposition of formyl, ethyl and propyl radical inter-mediates leads to the generation of _H atoms and _CH3 radicals, which are responsible, via H-atom abstraction, for 90% of fuel consumption, at 1200 K. 0 20 40 60 80 100 0 2 4 6 8 10 12 14 C₃H₇CHO CO C/C 0 [% ] C₂H₂ C₂H₆ C/C 0 [% ] T (K) 0 20 40 60 80 1000 11 00 1200 130 0 1400 1000 1100 1200 1300 1400 0 2 4 6 8 10 C₂H₄ CH₄ C₃H₈ C 3H6 T (K)

Fig. 12. Predicted and experimental concentration profiles from shock tube pyrolysis of 3% n-butanal in argon (s= 2.0 ms). Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

Figure 14 compares the species predictions by both mecha-nisms to experimental data for 3% n-pentanal pyrolysis in argon. Both of the models are able to reproduce the fuel conversion, indi-cating that n-pentanal starts to decompose at temperatures of 1000–1050 K. Both of the mechanisms reproduce the relative importance of all species. Within the experimental uncertainty, the two kinetic schemes satisfactorily reproduce the minor hydro-carbon species.

Experimentally, the three aldehydes seem to behave very clo-sely for temperatures up to 1200 K. Propanal shows the highest conversion rate for T > 1200 K, while no big differences are high-lighted between n-pentanal and n-butanal. As schematically shown in Fig. 5, this can be justified by the fact that the H-abstrac-tion from the highly reactive

a

and b positions of propanal results in a very high production of H

radical, via b-scission reactions. Both the mechanisms predict very similar reactivity up to 1200 K, with butanal showing the highest conversion rate for higher temperatures.

Both mechanisms agree with the pyrolytic experiments, indi-cating that the core pyrolytic kinetics and thermodynamics are of reasonable accuracy. Particularly, both the models satisfactorily reproduce fuel and methane profiles for the three aldehydes, indi-cating that rate constants for unimolecular decomposition (initia-tion step), abstraction by hydrogen and methyl radical (propagation steps), and methyl-methyl radical recombination (termination step) are well-constrained.

5.2. Ignition delay times in shock tubes 5.2.1. Propanal

Ignition delay times measurements at 1 and 3 atm reflected pressures made in this study are shown in Fig. 15 together with modeling predictions. Experiments were carried out using 1% fuel in O2/argon as described in Table 2 (Section 2.2). While the mech-anisms are in good or reasonable agreement with the experimental data at 3 atm under all investigated conditions, larger deviation are 0 20 40 60 80 100 0 20 40 60 80 100 1000 1100 1200 1300 1400 0 1 2 3 4 5 6 1000 1100 1200 1300 1400 0 5 10 15 C₄H₉CHO CO C/ C0 [% ] C₂H₄ CH₄ C₂H₆ C₂H₂ C/ C0 [% ] T (K) C₃H₆ C₃H₈ T (K)

Fig. 14. Predicted and experimental concentration profiles from shock tube pyrolysis of 3% n-pentanal in argon (s= 2.3 ms). Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

5,5 6,0 6,5 7,0 7,5 8,0 8,5 9,0 10 100 1000 6,0 6,5 7,0 7,5 8,0 8,5 9,0 10 100 1000 φ=2.0 φ=1.0 φ=0.5 Ign it ion De la y Ti me [ μ s] 10000/T (1/K) p=1 atm p=3 atm φ=2.0 φ=1.0 φ=0.5 10000/T (1/K)

Fig. 15. Predicted and experimental ignition delay times of 1% propanal in O2/Ar mixtures. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

observed at atmospheric pressure. In this case, both of the mecha-nisms predict higher apparent activation energies, particularly at stoichiometric and fuel-lean conditions.

Figure 16 compares model predictions and ignition delay times measured by Akih-Kumgeh and Bergthorson [51]. In line with the previous observations of Fig. 15, while good agreement is obtained

at 12 atm pressure, at atmospheric pressure the activation energy is over-predicted by both mechanisms. The NUIG mechanism shows better agreement with the data of Akih-Kumgeh, primarily for stoichiometric atmospheric conditions.

Figure 17 shows a global rate of production analysis at stoichi-ometric conditions, atmospheric pressure, and at temperatures of

6,0 6,5 7,0 7,5 8,0 8,5 9,0 10 100 1000 7,0 7,5 8,0 8,5 9,0 10 100 1000 φ = 1.0 φ = 0.5

Ignition Delay Time [

μ s] 10000/T (1/K) p=1 atm p=12 atm φ = 1.0 φ = 0.5 10000/T (1/K)

Fig. 16. Predicted and experimental ignition delay times of O2/Ar mixtures containing 1.25% propanal [51]. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

Fig. 17. Rate of production analysis for 1% propanal in O2/Ar mixtures atu= 1.0, p = 1 atm. NUIG mechanism, T = 1150 K (italic) and 1620 K (bold).

H+O2<=>O+OH H2O2(+M)<=>OH+OH(+M) C2H5CHO+HO2<=>C2H5CO+H2O2C2H4+OH<=>C2H3+H2O CH3+HO2<=>CH3O+OH C2H3+O2<=>CH2CHO+O C2H4+H(+M)<=>C2H5(+M)C2H5CHO<=>C2H5+HCO HO2+H<=>OH+OH HCO+M<=>H+CO+MOH+H2<=>H+H2O C2H2+O<=>HCCO+H HCO+OH<=>CO+H2OHCO+H<=>CO+H2 C2H3+H<=>H2CC+H2 C2H2+H(+M)<=>C2H3(+M)C2H3+H<=>C2H2+H2 CH3+H(+M)<=>CH4(+M)CH3+CH3<=>H+C2H5 C2H5CHO+H<=>C2H5CO+H2C2H3+O2<=>CH2O+HCO C2H5CHO+H<=>C2H4CHO-1+H2H2+O2<=>H+HO2 HCO+O2<=>CO+HO2 CH3+HO2<=>CH4+O2 C2H5CHO+H<=>C2H4CHO-2+H2HO2+HO2<=>H2O2+O2 HO2+OH<=>H2O+O2 -0,3 -0,2 -0,1 0,0 0,1 0,2 S_i x 0.5 Si Reaction i 1150 K 1620 K

Fig. 18. Sensitivity coefficients of ignition delay times to rate constants for 1% propanal in O2/Ar mixtures atu= 1.0, p = 1 atm, 1150 K (black bars) and 1620 K (gray bars). NUIG mechanism.

1150 K and 1620 K for the NUIG mechanism. At 1620 K, the unimo-lecular decomposition pathways contribute 8% to fuel consump-tion, being negligible at 1150 K. The fuel decomposition mostly occurs through the

a

channel, either via direct H-atom abstractions or due to isomerization of b and

c

radicals, with a relevant forma-tion of _C2H5 radicals. At 1620 K ethyl radicals mostly decompose to ethylene and H_ atoms promoting reactivity, while at 1150 K ethyl radicals can also react with molecular oxygen producing H _O2 rad-icals, thus inhibiting the system.

Figure 18 shows the sensitivity coefficients of ignition delay times to rate constants in the NUIG mechanism. Sensitivity coeffi-cients were calculated for each reaction via a brute force method, where a negative coefficient indicates a reaction promoting reac-tivity, i.e. decreases ignition delay times. Increasing the rates of H-atom abstraction reactions by H_ atoms decreases reactivity mak-ing ignition delay times longer. This is because by reacting with any species but O2a H_ atom is removed from the system to form H2, rather than reacting with O2 to generate €O atoms and _OH rad-icals which is the predominant chain-branching process for high temperature combustion. H-atom abstraction by H _O2 radicals on the

a

site appears as a sensitive parameter at 1150 K, due to the subsequent decomposition of H2O2 to generate two _OH radicals, which is highlighted as a promoting reaction. The importance of radical species such as formyl, methyl, vinyl and ethyl radicals is also highlighted, confirming the influence of reactions involving b-scission products such as ethylene and acrolein in the correct determination of propanal ignition. It is also of interest to observe the strong competition between vinyl radical decomposition through the third body reaction forming acetylene and _H atom, enhancing reactivity, and the bimolecular disproportionation reac-tion to form acetylene and H2, inhibiting reactivity.

5.2.2. n-Butanal

Figure 19 shows a comparison between experimental and cal-culated ignition delay times for mixture of 1% n-butanal in O2/ argon. Both of the mechanisms agree with the experimental data at the conditions tested, with the NUIG mechanism being consis-tently slower compared to the POLIMI mechanism, particularly at atmospheric and fuel rich conditions. The larger deviations from the experimental observations, mainly in terms of apparent activa-tion energies, can be observed for the POLIMI and NUIG mecha-nisms at 1 atm and

u

= 0.5.

Figure 20 compares the mechanisms predictions with the experimental measurements of Davidson et al. [54], at two differ-ent dilutions. The NUIG mechanism reproduces the apparent acti-vation energy in both cases, while the POLIMI mechanism, slightly

5,5 6,0 6,5 7,0 7,5 8,0 8,5 10 100 1000 6,5 7,0 7,5 8,0 8,5 100 1000 p=1 atm 10000/T (1/K)

Ignition Delay Time [

μ s] φ=2.0 φ=1.0 φ=0.5 p=3 atm 10000/T (1/K) φ=2.0 φ=1.0 φ=0.5

Fig. 19. Predicted and experimental ignition delay times of O2/Ar mixtures containing 1% n-butanal. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

6,5 7,0 7,5 8,0 8,5

100 1000

p =1.7 atm

Ignition Delay Time [

μ

s]

φ=2.0

φ=1.0

10000/T (1/K)

Fig. 20. Predicted and experimental ignition delay times of O2/Ar mixtures containing 1% n-butanal [54]. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

faster than NUIG, over-predicts the activation energies, but is still in reasonable agreement with experiments.

Figure 21 compares all three sets of experimental data (our experiments at 1 atm, Zhang et al. [56] at 1.3 atm and Davidson et al. [54] at 1.7 atm). Following the approach of Davidson et al. [54], all of the data have been scaled at 1.7 atm, assuming a scaling factor pexp/1.70.52. The experimental data agree under stoichiome-tric conditions, while the measurements of Zhang are notably fas-ter under fuel-rich conditions. Figure 22 compares the experimental data from Zhang et al. [56] with model predictions. Both of the mechanisms tend to over-predict ignition delay times, particularly at 1.3 atm. The POLIMI mechanism reproduces these experiments more closely than the NUIG mechanism, which over-predicts the ignition delay times, particularly at atmospheric pressure. However, on the basis of the observations of Fig. 21, the performances of the two mechanisms are considered to be in sat-isfactory agreement with the experimental data.

Figure 23 shows a reaction path analysis carried out for fuel-lean (

u

= 0.5) and fuel-rich (

u

= 2.0) mixtures at 3 atm and 1320 K, with the NUIG mechanism. The unimolecular decomposi-tion reactions accounts for 12.6% and 7.5% of fuel consumption under fuel-rich and fuel-lean conditions, respectively. H-atom abstraction from the

a

site is an important decomposition channel (27.0%). The b and

c

sites show similar selectivities (20–30%), with b favored at fuel-rich conditions and

c

at fuel-lean conditions.

This can be explained on the basis of a higher production of _OH radicals through the chain-branching reaction H_ þ O2 $ O€ þ _OH, due to the higher concentration of oxygen in the system. _OH radi-cals are then more likely to abstract on the

c

site relative to the b site.

Figure 24 presents sensitivity coefficients of ignition delay times to rate constants for NUIG mechanism at the same condi-tions shown in Fig. 23, in order to identify the reactions controlling the auto-ignition behavior. Chain initiation reactions generating

highly reactive radicals promote the system’s reactivity at both conditions. Particularly, the breaking of the weak CbACcbond

pro-ducing an ethyl radical and the breaking of the CaACbbond pro-ducing a formyl radical exhibit some sensitivity. Decreasing the oxygen concentration largely enhances the importance of small unsaturated species such as acetylene, ethylene, propylene and the parent radicals. H-atom abstraction reaction by _H atoms reduces the reactivity due to the competition with the branching reaction _H þ O2$ €O þ _OH. Furthermore, H-atom abstractions from

10 100 1000 6,5 7,0 7,5 8,0 8,5 5,5 6,0 6,5 7,0 7,5 8,0 10 100 1000 This Work Davidson et al Zhang et al. .

Ignition Delay Time [

μ s] 10000/T (1/K) p=1.7 atm, φ=1.0 _{p=1.7 atm, φ=2.0} This Work Davidson et al. Zhang et al. 10000/T (1/K)

Fig. 21. Comparison amongst experimental ignition delay times for n-butanal-O2/Ar mixtures, scaled at 1.7 atm. This work: 1% fuel at 1 atm (squares); Zhang et al. [56]: 1.2%fuel at 1.3 atm (triangles); Davidson et al. (circles) [54]: 1% fuel at 1.7 atm.

100 1000 100 1000 0,65 0,70 0,75 0,80 0,85 0,65 0,70 0,75 0,80 0,85 0,65 0,70 0,75 0,80 0,85 100 1000 p =1.3 atm 1.2% n-butanal Ignition De lay T ime [ μ s] φ=0.5 φ=1.0 φ=2.0 1000/T (1/K) p =5 atm, 0.6% n-butanal 1000/T (1/K) p =10 atm, 0.6% n-butanal 1000/T (1/K)

Fig. 22. Predicted and experimental ignition delay times of n-butanal in O2/Ar mixtures [56]. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).

the

a

site increase the ignition delay times due to the formation of 1-propyl radical, which mainly decomposes to produce methyl radicals. Analyses carried out for POLIMI mechanisms highlighted the same classes of fuel specific reactions as sensitive to the ignition delay time determination.

5.2.3. n-Pentanal

Experimental and calculated ignition delay times for 1% n-pen-tanal mixtures in O2/Ar are reported in Fig. 25 for conditions as described in Table 2. The NUIG mechanism captures the experi-mental data at every condition tested. The POLIMI mechanism tends to predict shorter ignition delay times for fuel-rich condi-tions particularly at atmospheric pressure. At 1850 K, the ignition delay time is under-predicted by a factor of three. Again, the model over-predicts the apparent activation energy. At 3 atm both of the models over-estimate the ignition delay time at fuel-lean condi-tions for temperatures below 1100 K. Low-temperature chemistry, outside the scope of this work, might have an effect at these conditions.

Figure 26 presents a rate of production analysis carried out for the POLIMI mechanism at intermediate conditions (

u

= 1.0, p = 2 atm, T = 1300 K). Unimolecular decomposition reactions account for less than 3.0% of n-pentanal consumption. Decomposi-tion occurs mainly via H-atom abstraction from the

a

position (54%), followed by abstraction from the

c

and d (15%) positions. At the conditions investigated 75% of the

e

-radical ( _C4H8CHO-4) isomerizes to the

a

-radical (n-C4H9 _CO), which is assumed to elim-inate CO (90%) and/or decompose to n-propyl and ketene (10%). As already discussed in Section 4.2, the POLIMI lumped mechanism only accounts for the

e

-radical, the rate of production is simply obtained through a de-lumping procedure.

Figure 27 shows the sensitivity analyses of ignition delay times to rate constants for the NUIG and POLIMI mechanisms, at

u

= 1.0, p = 2 atm, and T = 1300 K. With regards to fuel specific reactions,

H+O2<=>O+OH CH3+HO2<=>CH3O+OH NC3H7CHO<=>C2H5+CH2CHOC2H4+OH<=>C2H3+H2O NC3H7CHO<=>NC3H7+HCOC2H3+O2<=>CH2CHO+O CH2CHO(+M)<=>CH2CO+H(+M)CH3+O2<=>CH2O+OH H2O2(+M)<=>OH+OH(+M) C3H5-A+HO2<=>C3H5O+OHCH2CO+H<=>HCCO+H2 HCCO+O2=>CO2+CO+HHCO+M<=>H+CO+M HCO+O2<=>CO+HO2 CH2CO+H<=>CH3+CO C3H6+H<=>C3H5-A+H2 C2H2+H(+M)<=>C2H3(+M)CH4+OH<=>CH3+H2O CH3+H(+M)<=>CH4(+M) NC3H7CHO+H<=>C3H6CHO-2+H2 NC3H7CHO+H<=>C3H6CHO-3+H2H2+O2<=>H+HO2 NC3H7CHO+OH<=>NC3H7CO+H2OC3H6+OH<=>C3H5-A+H2O CH2CHO(+M)<=>CH3+CO(+M)HO2+OH<=>H2O+O2 NC3H7CHO+H<=>NC3H7CO+H2CH3+CH3(+M)<=>C2H6(+M) NC3H7CHO+H<=>C3H6CHO-1+H2CH3+HO2<=>CH4+O2 -0,4 Re ac tion i H H O O O O) H) H2 H M2 O2 ) O) 2 2 2 O O) 2 2) 2 2 -0,3 -0,2 -0,1 0,0 0,1 S_i x 0.5 Si φ=0.5 φ=2.0

Fig. 24. Sensitivity coefficients of ignition delay times to rate constants for 1% n-butanal in O2/Ar mixtures at p = 3 atm, T = 1320 K atu= 0.5 (black bars) andu= 2.0 (gray bars). 5,5 6,0 6,5 7,0 7,5 8,0 8,5 10 100 1000 6,0 6,5 7,0 7,5 8,0 8,5 10 100 1000 φ = 2.0 φ = 1.0 φ = 0.5 Ign ition D ela y Tim e [ μ s] p= 1 atm 10000/T [1/K] 10000/T [1/K] φ = 2.0 φ = 1.0 φ = 0.5 p= 3 atm

Fig. 25. Predicted and experimental ignition delay times of O2/Ar mixtures containing 1% n-pentanal. Experiments (symbols), POLIMI mechanism (solid lines) and NUIG mechanism (dashed lines).